Rubbish unloading roller shutter door intelligent monitoring method and system based on multi-source sensing fusion

Abstract

The invention relates to the technical field of roller shutter door control, and discloses a garbage discharging roller shutter door intelligent monitoring method and system based on multi-source sensing fusion. When the roller shutter door receives an opening or closing instruction to start running, a high-definition camera is started to obtain a real-time image, the real-time image and a background model are analyzed to recognize whether foreign matter exists or not, if yes, early warning is conducted, and if not, the next stage is started; in the normal advancing process of the door body, the straightness of the track and whether the door curtain deforms or not are judged, and if obvious deformation exists, early warning is conducted; otherwise, according to a vibration signal and a pressure signal during operation of the roller shutter door body, feature extraction is conducted on the operation state of the roller shutter door to obtain state feature parameters, finally, comprehensive analysis is conducted according to the state feature parameters to output an operation index, and if the operation index exceeds a threshold value, it is judged that remarkable abnormity or potential faults exist in the roller shutter door. Otherwise, judging that the equipment state is normal or acceptable, recovering to the low-power-consumption standby state after the monitoring record is completed, and waiting for the triggering of the next operation instruction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rolling shutter door control, in particular to an intelligent monitoring method for garbage unloading rolling shutter door based on multi-source sensor fusion. BACKGROUND

[0002] The garbage unloading rolling shutter door is a special industrial door installed at the unloading port of garbage transfer stations, incineration plants or recycling centers, etc. Its core role is to efficiently seal and isolate during garbage truck unloading, prevent odor from overflowing and improve work efficiency. As a key equipment for garbage unloading at the transfer station, the rolling shutter door is subjected to multiple challenges such as corrosive medium, high humidity environment, frequent start-stop and vehicle collision risk for a long time. In such harsh conditions, the door body guide rail is prone to fouling or deformation, and the driving mechanism is prone to overload and wear, resulting in running jam. At the same time, the door curtain itself may be locally deformed due to external force or fatigue, which destroys the overall sealing and balance. More seriously, during the operation of the door body, foreign matter intrusion (such as incomplete vehicle departure and garbage accumulation) can easily cause serious mechanical damage or safety accidents. Therefore, it is necessary to monitor the garbage unloading rolling shutter door. At present, the garbage unloading rolling shutter door monitoring is realized by controlling the opening and closing state through radar, geomagnetic induction, etc. to realize basic automation, but the fault monitoring is usually checked by the staff during operation to find out. This monitoring method is backward, highly dependent on manual work, poor in reliability, and unable to timely discover faults and handle them. SUMMARY

[0003] In view of the problems in the related art, the present application provides an intelligent monitoring method for garbage unloading rolling shutter door based on multi-source sensor fusion to overcome the technical problems existing in the prior art.

[0004] To solve the technical problems, the present application is realized by the following technical solutions: On the one hand, the present application provides an intelligent monitoring method for garbage unloading rolling shutter door based on multi-source sensor fusion, which specifically includes: Step one, when the rolling shutter door receives an opening or closing instruction to start running, the door body image, vibration signal and pressure signal are collected; Step two, an alien object detection method based on background modeling and deep learning fusion is used to determine whether there is an obstacle on the door body running path. If an alien object is detected, the door body movement is paused and an alarm is issued. If not, step three is executed; Step three, the door body deformation is monitored by an infrared distance sensor array. The real-time distance-height curve is compared with the pre-stored ideal reference curve to calculate the deformation variable value. If the deformation variable value exceeds the deformation threshold, the door body movement is paused and a deformation warning is generated. If not, step four is executed and the variable value is transmitted to step five; Step four, based on the real-time acquisition of vibration information and pressure information, respectively, feature extraction is carried out to obtain vibration characteristic value and pressure characteristic value, and they are transmitted to step five; Step five, the deformation value, vibration characteristic value and pressure characteristic value are quantified by linear formula to obtain running index, if the running index is greater than the running threshold, the running maintenance early warning is generated, otherwise no operation is needed, and it returns to low-power standby state, waiting for the trigger of next running instruction.

[0005] Preferably, the foreign matter detection method based on background modeling and deep learning fusion: The background model is preset, the stereo image is extracted, the absolute difference between each pixel of the door body image and the background model is calculated in real time to obtain the difference value of each pixel, and then the difference values of each pixel are combined to form a difference image (x, y); The difference image (x, y) is binarized: if the difference value is greater than the preset threshold, the pixel coordinates of the difference image corresponding to the difference value are assigned a value of 1, otherwise a value of 0; then the binarized difference image (x, y) is subjected to open operation, and the white connected region is marked as a candidate foreign matter region; the trained target monitoring neural network is used to search and identify the door body image and the candidate foreign matter region, and the foreign matter region boundary box and the confidence are output; The confidence threshold is preset, if the confidence is greater than the preset confidence threshold, the foreign matter early warning is generated, and the roller shutter door movement is paused.

[0006] Preferably, the background model construction process: The background model refers to a static reference picture, which represents the normal appearance of the door body region in the absence of foreign matter. In the absence of foreign matter, a plurality of images are collected to establish a background model, and the color or gray value of each pixel point in the image sequence is statistically analyzed to obtain a background model image with the same size as the real-time image, wherein each pixel point stores the color or gray value.

[0007] Preferably, the real-time distance-height curve is compared with the pre-stored ideal reference curve to calculate the deformation value: 4-1, a plurality of measuring points are arranged at equal intervals from top to bottom on both sides of the roller shutter door, and each measuring point is provided with an infrared distance measuring sensor. Each sensor continuously measures the actual distance from the sensor to the door body surface at a fixed sampling frequency, which is recorded as di, i is the index of the infrared distance measuring sensor, that is, the index of the measuring point; thus the actual curve of the actual distance di of each sensor i with the change of the door body height is recorded as di (H), H is the door body height; 4-2, each sensor i corresponds to an ideal curve of the distance Di changing with the door body height H in the normal state; 4-3. For each sensor i, extract its corresponding ideal curve, and compare the current actual curve with the ideal curve to obtain the deformation value; 4-4 Thus, each time the roller shutter door moves from one extreme to another, it corresponds to a deformation value. A deformation threshold is preset. If the deformation value is greater than the deformation threshold, the movement of the roller shutter door is paused and a roller shutter door deformation warning is generated.

[0008] Preferably, the deformation values ​​are obtained by overlapping comparison: Extract any one of the door heights Hj, and calculate the absolute deviation Δdj between the two curves at the door height Hj. Δdj = |di(Hj) - Di(Hj)|, where j is the index of any door height, j = 1, 2, 3, ..., N, where N is a positive integer and N represents the total number of different door heights. From this, the deviation sequence {Δdj} can be obtained. Using formula To calculate the root mean square error R, we treat the ideal curve and the actual curve as two variable sequences and calculate the Pearson correlation coefficient between them to obtain the correlation coefficient r. Specifically: ,in This represents the average distance of the actual curve at different gate heights. This represents the average distance of the ideal curve at different door heights; The root mean square error R and the correlation coefficient r are expressed by the formula The deformation values ​​were calculated. Rmax is the preset maximum allowable root mean square error, and λ1 and λ2 are the weighting coefficients of the root mean square error and the correlation coefficient, respectively.

[0009] Preferred vibration feature value extraction process: The vibration signal is extracted and denoted as A(t), and it is plotted as a vibration time-domain spectrum, where t is the time index and A(t) represents the amplitude at time t; the vibration time-domain spectrum of the vibration signal A(t) is analyzed: (1) by formula Calculate the root mean square value of the vibration signal , which characterizes the overall energy level of the vibration signal, and T represents the total time; (2) through the formula Calculate the waveform value of the vibration time-domain spectrum It represents the proportion of the impact component in the vibration signal relative to the overall average level; The root mean square value of the vibration time-domain spectrum and waveform value pass The nonlinear enhancement model calculates vibration characteristic values ​​to measure the vibration state of the roller shutter door during its movement, where η1 and η2 are the root mean square values, respectively. and waveform value The weighting coefficients.

[0010] Preferred pressure feature value extraction process: The extracted pressure signal is denoted as Yk, where k is the index of the pressure sensor at the bottom sealing strip of the door, k=1,2,3……K; the pressure signal Yk at the same time t is obtained through the formula The uniformity of pressure distribution is quantified to obtain the coefficient of variation P(t) between pressure sensors, where Let be the average pressure of K sensors at time t; from this, the coefficient of variation P(t) at each time can be obtained, and a time-domain spectrum of variation can be constructed. Spectral analysis of the time-domain spectrum of variation is then performed: (1) Through the formula Calculate the root mean square value of the vibration signal , which characterizes the degree of continuous uneven pressure; (2) through the formula Calculate the waveform values ​​of the time-domain spectrogram of the variation. It characterizes the impact characteristics and stability of the pressure variation sequence; The root mean square value of the variation time-domain spectrum and waveform value pass The nonlinear enhancement model calculates the pressure characteristic values ​​to measure the vibration state of the roller shutter door during its movement, where η3 and η4 are the root mean square values, respectively. and waveform value The weighting coefficients.

[0011] On the other hand, the present invention provides an intelligent monitoring system for waste unloading roller shutter doors based on multi-source sensor fusion, specifically including: a sensing module, a foreign object monitoring module, a deformation monitoring module, a status monitoring module, and an intelligent decision-making module; The sensing module is used to collect images of the door body, as well as vibration and pressure signals, when the roller shutter door starts running after receiving an open or close command. The foreign object detection module, based on a foreign object detection method that combines background modeling and deep learning, determines whether there are obstacles in the door's running path. If a foreign object is detected, the door movement is paused and an alarm is issued; otherwise, the deformation detection module is executed. The deformation monitoring module monitors the deformation of the gate through an infrared ranging sensor array. It compares the real-time distance-height curve with the pre-stored ideal reference curve to calculate the deformation value. If the deformation value exceeds the deformation threshold, the gate operation is paused and a deformation warning is generated; otherwise, the status monitoring module is executed and the variable value is transmitted to the status monitoring module. The condition monitoring module extracts features based on real-time acquired vibration and pressure information to obtain vibration feature values ​​and pressure feature values, and then transmits them to the intelligent decision-making module. The intelligent decision module obtains an operation index by comprehensively quantifying the deformation variable value, the vibration characteristic value and the pressure characteristic value through a linear formula, generates an operation maintenance early warning if the operation index is greater than an operation threshold, and does not need to perform any operation if not, and returns to a low-power standby state, waiting for triggering of a next operation instruction.

[0012] The present application has the following advantages: 1. The foreign matter detection method fusing background modeling and deep learning establishes a precise background model as a benchmark, efficiently extracts candidate foreign matter regions by using difference calculation, morphological processing and connected region analysis, and then combines a trained and improved target detection neural network to realize precise recognition and confidence evaluation, so as to realize active early warning of safety protection, reliably identify various foreign matters such as personnel, tools and large garbage, and timely suspend the movement of the door body and alarm before the foreign matter invasion may cause equipment damage or personal injury.

[0013] 2. The actual curve is collected on the complete operation track of the door body by the ranging array, and the actual curve is compared with the ideal curve to calculate the deviation sequence and the root mean square error, and then the morphological consistency is evaluated by using the Pearson correlation coefficient, and finally the root mean square error and the Pearson correlation coefficient are combined into a deformation variable according to the weight, so that the deformation variable of the roller shutter door body can be accurately monitored, and timely warning can be made to improve the early identification and reduce the ability of false stop.

[0014] 3. The vibration signal and the sealing strip pressure signal are analyzed in the time domain to output the vibration characteristic value representing the vibration state of the roller shutter door body; at the same time, the pressure signal is calculated to obtain the pressure variation coefficient, and the pressure variation coefficient at each time is analyzed in the time domain to output the pressure characteristic value representing the pressure state of the roller shutter door body; the running state of the roller shutter door body can be accurately quantified, which lays a foundation for state monitoring.

[0015] 4. The deformation variable value representing the structure state, the vibration characteristic value reflecting the mechanical state and the pressure characteristic value reflecting the sealing performance are weighted and fused into an operation index to realize overall evaluation and intelligent decision of the health state of the roller shutter door, overcome the limitations of single parameter monitoring, and comprehensively reflect the mechanical structure integrity, driving system running stability and sealing performance of the equipment, so as to provide accurate and reliable basis for predictive maintenance, significantly improve the operation and maintenance efficiency and reduce the life cycle cost.

[0016] Of course, any product implementing the present application does not necessarily need to achieve all the advantages described above. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the description of the embodiments, obviously, the drawings in the following description are only some of the embodiments of the present application, and for those skilled in the art, the drawings can also obtain the drawings without paying creative labor.

[0018] Figure 1 The present application provides a flow chart of a garbage unloading roller shutter door intelligent monitoring method based on multi-source sensor fusion. Figure 2 The present application provides a module diagram of a garbage unloading roller shutter door intelligent monitoring system based on multi-source sensor fusion. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application, obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without paying creative labor belong to the scope of protection of the present application.

[0020] At present, the garbage unloading roller shutter door monitoring is realized by controlling the opening and closing state through radar, geomagnetic induction and the like, realizing the basic automation, but the fault monitoring is usually checked and found by the staff during operation, this monitoring mode is backward, highly dependent on manual operation, poor reliability, and unable to timely find faults and timely handle.

[0021] To solve the problems mentioned in the background art, as shown in Figure 1 The embodiments of the present application provide a garbage unloading roller shutter door intelligent monitoring method based on multi-source sensor fusion, specifically including the following steps: Step one, through the communication connection with each sensor installed at the garbage unloading roller shutter door, the state information of the garbage unloading roller shutter door is acquired in all directions and multidimensionally, and the specific state information includes vibration signal, door body image and pressure signal; The specific sensor arrangement is as follows: A vibration sensor is installed at the drive motor housing to collect vibration signals; A corrosion-resistant high-definition camera is installed on the wall corner or top beam above the one side of the roller shutter door door hole to collect high-definition images; ensure that the camera and the door body plane form a certain angle (usually 15-45°), obliquely downward view the entire roller shutter door body running area, the entire process from the top completely closed to the bottom completely opened can be observed completely, and the intrusion of foreign matters and macroscopic deformation can be captured; A pressure sensor is installed at the bottom seal strip of the roller shutter door body to collect the pressure at the bottom of the door body; In practical applications, since the majority of the roller shutter door is in a static state (constant opening state or constant closing state), periodic activation of the sensor will generate a large amount of useless data, greatly wasting storage space, computing resources and network bandwidth. Therefore, the application scenario of the present application is to activate the sensor to collect information before and during the activation of the roller shutter door. Specifically, when the roller shutter door is activated (opened or closed), the sensor is activated to collect information until the roller shutter door completes the action and returns to the standby state.

[0022] Step two, a preset background model is provided, the background model refers to a static reference picture, which represents the normal appearance of the door body area in the absence of foreign matter. In practical applications, the background modeling process is as follows: in the absence of foreign matter, multiple frames of images are collected to establish a background model. The color or gray value of each pixel point in the image sequence is statistically analyzed to obtain a background model image with the same size as the real-time image, wherein each pixel point stores the typical color or gray value of the position in the normal state; The door body image is extracted, and the absolute difference between each pixel of the door body image and the background model is calculated in real time. The difference image (x, y) = |real-time image (x, y) - background model (x, y)|, (x, y) is the pixel coordinate. The difference value of each pixel position is calculated by difference value calculation to obtain the difference value of each pixel. Then the difference values of each pixel are combined to form the difference image (x, y). The difference image (x, y) is binarized: if the difference value is greater than the preset threshold, the pixel coordinate of the difference image corresponding to the difference value is assigned a value of 1, otherwise a value of 0. Then, the binarized difference image (x, y) is subjected to an open operation to eliminate noise spots and connect the broken foreground area. Then, the white connected region (i.e. the pixel block with a value of 1) in the difference image is identified and marked as a candidate foreign matter area. Finally, the trained target monitoring neural network is used to search and identify the original stereo image and the candidate foreign matter area, and the foreign matter area boundary box and the confidence are output. The confidence threshold is set based on the balance strategy of the precision and recall rate in the performance evaluation of the classification model. The core principle is to achieve the optimal balance between false positives and false negatives in engineering. Specifically, by selecting the critical point with a higher score or meeting the business tolerance on the performance curve of the model on the validation set: usually, the confidence level (such as 0.7-0.9) that can cover most of the real threats is selected under the premise of ensuring the accuracy of foreign matter identification (such as > 95%). The threshold needs to be adjusted according to the actual scene. In the case of high safety requirements, the threshold is appropriately reduced to improve the sensitivity. In the case of serious false interference, the threshold is increased to ensure reliability. The final goal is to produce a stable response to real foreign matter while minimizing false shutdowns caused by environmental interference. If the confidence is greater than the preset confidence threshold, it indicates that there is a risk of foreign matter, and a foreign matter warning is generated, and the roller shutter door movement is suspended. Otherwise, step three is performed. The foreign matter detection method fuses background modeling and deep learning to establish a precise background model as a benchmark, uses difference calculation, morphological processing and connected region analysis to efficiently extract candidate foreign matter regions, and then combines a trained and improved target detection neural network to perform accurate recognition and confidence evaluation, thereby achieving active early warning for safety protection.

[0023] Step three, on both sides of the roller shutter door hole, a plurality of measuring points are arranged at equal intervals from top to bottom, and each measuring point is provided with an infrared distance sensor. These infrared distance sensors form a measurement array. When the roller shutter door runs uniformly from one limit position to another limit position, i.e. fully open or fully closed, each sensor continuously measures the actual distance from the sensor to the door surface at a fixed sampling frequency, denoted as di, i is the index of the infrared distance sensor, i.e. the index of the measuring point. Thus, the actual curve of the actual distance di of each sensor i with the change of the door height H is obtained, denoted as di(H), H is the door height; A preset distance Di of each sensor i corresponding to an ideal curve of the distance Di with the change of the door height H in a normal state is provided. The specific collection process is as follows: when it is determined that the roller shutter door is in a normal state (i.e. the track is straight and the door curtain is not deformed), the roller shutter door runs uniformly from one limit position to another limit position, is paused at multiple different heights of the door body, and static measurement is performed to obtain the relationship between the distance Di of the sensor i in the normal state and the door height H, and the ideal curve of the distance Di with the change of the door height H is drawn, denoted as Di(H). In actual application, no matter whether the door is opened or closed, the distance of each sensor when the door body is at a certain height is the same. Therefore, in an ideal case, the ideal curve of the distance Di with the change of the door height H is the same whether the door is opened or closed. For each sensor i, the corresponding ideal curve is extracted, and the current actual curve is compared with the ideal curve. When the comparison is made, it is necessary to ensure that the two curves are compared at the same height sequence (i.e. the same door height). Any one of the door heights Hj is extracted, and the absolute deviation dj of the two curves at the door height Hj is calculated, dj=|di(Hj)-Di(Hj)|, j is any door height index, j=1, 2, 3, …, N, N is a positive integer, and N represents the total number of different door heights. Thus, the deviation sequence {dj} is obtained, which reflects the offset of the door body relative to the reference state at each specific position. Then, the formula The root mean square error R is calculated, which is more sensitive to larger deviations in the deviation sequence and can better capture significant abnormal points in the curve; The ideal curve and the actual curve are regarded as two variable sequences, and the Pearson correlation coefficient between the two is calculated to obtain a correlation coefficient r, specifically: wherein is the distance mean of the actual curve at different door body heights, represents the distance mean of the ideal curve at different door body heights; it can be known from the formula that r takes a value between -1 and 1, when r≈1, the shapes of the two curves are highly consistent, when r≈0, the shapes of the two curves are not correlated, indicating that there is a serious deformation, and when r≈-1, the shapes of the two curves are completely opposite, indicating that there is a systematic reverse deformation; The root mean square error R and the correlation coefficient r are calculated by the formula to obtain a deformation variable value , Rmax is a preset maximum allowed root mean square error, which is usually the root mean square error with the highest occurrence frequency in actual application; λ1 and λ2 are weight coefficients of the root mean square error and the correlation coefficient respectively, which are taken as 0.6 and 0.4 respectively by those skilled in the art in actual application, and the specific values are adjusted according to actual conditions; Thus, each time the roller shutter door body runs from one extreme to the other (i.e. opening or closing), a deformation variable value is correspondingly obtained, and a preset deformation threshold is set, which is based on the principle of combining the statistical characteristics of historical deformation variable values of the equipment in the normal state with the engineering safety margin. Specifically, by analyzing the distribution range of a large number of deformation variable values in normal operation (usually taking the mean value plus two to three times the standard deviation as the threshold reference), it is ensured that the threshold can effectively distinguish between normal fluctuations and abnormal deformation. At the same time, the threshold is fine-tuned in combination with the technical specifications of the equipment manufacturer and the safety requirements of the actual application scene, so as to avoid false positives and reliably warn of significant deformation that may cause running jamming, sealing failure or structural damage. If the deformation variable value is greater than the deformation threshold, it indicates that the deformation of the roller shutter door body is obvious and needs to be repaired in time, then the roller shutter door movement is paused, and a roller shutter door deformation warning is generated to notify the maintenance personnel to repair in time. Otherwise, the deformation variable value is transmitted to step five; By collecting the actual curve on the complete running track of the door body through the ranging array, and performing point-by-point difference with the ideal curve, calculating the deviation sequence and the root mean square error, and then evaluating the shape consistency using the Pearson correlation coefficient, finally combining the root mean square error and the Pearson correlation coefficient according to the weight to obtain the deformation variable, the deformation variable of the roller shutter door body can be accurately monitored, and timely warning can be made to improve the ability of early identification and reduce false stops.

[0024] In step four, the vibration state and pressure state of the roller shutter door are respectively extracted based on the real-time acquired vibration information and pressure information of the roller shutter door to obtain state characteristic parameters, and the state characteristic parameters are transmitted to step five, wherein the state characteristic parameters include vibration characteristic values and pressure characteristic values, and specifically: Step 3-1, Vibration state feature extraction: The vibration signal is extracted and denoted as A(t), and it is plotted as a vibration time-domain spectrum, where t is the time index and A(t) represents the amplitude at time t; the vibration time-domain spectrum of the vibration signal A(t) is analyzed: (1) by formula Calculate the root mean square value of the vibration signal , which characterizes the overall energy level of the vibration signal, and T represents the total time; (2) through the formula Calculate the waveform value of the vibration time-domain spectrum The root mean square (RMS) of the vibration time-domain spectrum represents the proportion of the impact component in the vibration signal relative to the overall average level. In practical applications, when the waveform value is greater than 1.5, it is considered to have a significant impact component; the smaller the waveform value, the more obvious the characteristics of steady vibration, which is considered normal vibration. and waveform value pass The nonlinear enhancement model calculates vibration characteristic values ​​to measure the vibration state of the roller shutter door during its movement; the larger the value, the more obvious the vibration anomaly. Here, η1 and η2 are the root mean square values, respectively. and waveform value The weighting coefficients are set to 0.55 and 0.45 respectively by those skilled in the art, and can be modified according to changes in application scenario requirements; Step 3-2, Pressure State Feature Extraction: The extracted pressure signal is denoted as Yk, where k is the index of the pressure sensor at the bottom sealing strip of the door, k=1,2,3……K; the pressure signal Yk at the same time t is obtained through the formula The uniformity of pressure distribution is quantified to obtain the coefficient of variation P(t) between pressure sensors, where Let be the average pressure of K sensors at time t; from this, the coefficient of variation P(t) at each time can be obtained, and a time-domain spectrum of variation can be constructed. Spectral analysis of the time-domain spectrum of variation is then performed: (1) Through the formula Calculate the root mean square value of the vibration signal It characterizes the degree of continuous uneven pressure, and the larger the value, the greater the possibility of wear, deformation or local damage to the sealing strip; (2) through the formula Calculate the waveform values ​​of the time-domain spectrogram of the variation. It characterizes the impact characteristics and stability of the pressure variation sequence. A larger value indicates the presence of spikes and pulses in the variation coefficient sequence, suggesting sudden, intermittent, and uneven pressure distribution, potentially leading to problems such as momentary stagnation and unstable operation. The root mean square value of the variation time-domain spectrum... and waveform value pass The nonlinear enhancement model calculates the pressure characteristic value to measure the vibration state of the roller shutter during the movement process, and the greater the value, the more obvious the vibration anomaly; wherein η3 and η4 are the weight coefficients of the root mean square value, which are respectively taken as 0.55 and 0.45 by those skilled in the art, and can be modified according to the requirements of the application scene; and the weight coefficient of the waveform value , which are respectively taken as 0.55 and 0.45 by those skilled in the art, and can be modified according to the requirements of the application scene; The vibration signal and the sealing strip pressure signal are analyzed in the time domain spectrum to output the vibration characteristic value representing the vibration state of the roller shutter door body during operation. At the same time, the pressure signal is calculated to obtain the pressure variation coefficient, and the pressure variation coefficient at each time is analyzed in the time domain spectrum to output the pressure characteristic value representing the pressure state of the roller shutter door body. The running state of the roller shutter door body can be accurately quantified, which lays a foundation for state monitoring.

[0025] Step five, multi-dimensional running monitoring of the roller shutter door is carried out based on the state characteristic parameters and the deformation value to determine the state of the roller shutter door, and the corresponding monitoring result is generated, which is specifically: The deformation value , the vibration characteristic value and the pressure characteristic value are quantified by a linear formula to obtain the running index, and the specific linear formula is , wherein ω1, ω2 and ω3 are the weight coefficients of the deformation value , the vibration characteristic value and the pressure characteristic value , which are respectively taken as 0.4, 0.3 and 0.3 by those skilled in the art, and can be modified according to the requirements of the application scene; A preset running threshold is determined based on the statistical distribution of long-term running data of the equipment in a healthy state and the marginal effect allowed by the safe running of the equipment, which is specifically a critical value that can sensitively identify the comprehensive performance degradation and avoid excessive early warning, so as to ensure that the system triggers maintenance intervention in time when the overall state exceeds the normal fluctuation range, and avoid fault expansion; if the running index is greater than the running threshold, it indicates that the roller shutter door has significant abnormalities or potential faults, and needs to be maintained, then the running maintenance warning is generated; otherwise, it is in a normal or acceptable running state, and no operation is needed, and it returns to the low-power standby state and waits for the trigger of the next running instruction; By weighting and fusing the deformation value representing the structure state, the vibration characteristic value reflecting the mechanical state and the pressure characteristic value embodying the sealing performance into an operation index, the overall health state of the roller shutter door is evaluated and intelligent decision is realized, the limitations of single parameter monitoring are overcome, the mechanical structure integrity, the driving system running stability and the sealing performance of the equipment can be comprehensively reflected, thereby providing accurate and reliable basis for predictive maintenance, and the operation and maintenance efficiency is significantly improved and the whole life cycle cost is reduced.

[0026] The application is activated into a working state when the roller shutter door receives an open or close instruction to start running. First, the high-definition camera is started and the pre-established background model is called. Whether there is an obstacle on the running path of the door body is judged through real-time image difference calculation, binary processing and deep learning foreign object identification. If a high-confidence foreign object is detected, the door body movement is immediately paused and an alarm is issued. If the passing area is safe, the next stage is entered. In the normal running process of the door body, the infrared distance sensor array arranged on both sides of the door hole continuously measures the distance from the door body surface. The distance-height curve formed in real time is compared with the pre-stored ideal reference curve. The deformation value of the door body is comprehensively evaluated through calculation of the root mean square error and the correlation coefficient, which is used to judge whether the track is flat and whether the curtain is deformed. If there is obvious deformation, the door body running is stopped, and the roller shutter door deformation warning is triggered. Otherwise, the vibration signal and the pressure signal of the roller shutter door during running are used to extract the state characteristic parameters.

[0027] Finally, the deformation value, the vibration characteristic value and the pressure characteristic value are fused into a comprehensive operation index according to the preset weight. The index is compared with the operation threshold value. If the threshold value is exceeded, it is determined that the roller shutter door has significant abnormalities or potential faults, and a maintenance warning containing specific abnormal information is automatically generated to guide the maintenance personnel to implement accurate maintenance. If the threshold value is lower than the threshold value, it is determined that the equipment state is normal or acceptable. After completing the monitoring record, it returns to the low-power standby state and waits for the trigger of the next operation instruction. The whole process of the scheme realizes the full automation from state perception, abnormal diagnosis to maintenance decision, and changes the backward mode of relying on manual inspection.

[0028] As shown in Figure 2 The application embodiment provides a garbage unloading roller shutter door intelligent monitoring system based on multi-source sensing fusion. The system comprises a perception module, a foreign object monitoring module, a deformation monitoring module, a state monitoring module and an intelligent decision module. The perception module is used for collecting the door body image, the vibration signal and the pressure signal when the roller shutter door receives an open or close instruction to start running. The foreign matter monitoring module is based on a foreign matter detection method combining background modeling and deep learning to determine whether there is an obstacle on the running path of the door body. If foreign matter is detected, the door body movement is paused and an alarm is issued. If not, the deformation monitoring module is executed. The deformation monitoring module monitors the deformation of the door body through an infrared distance sensor array, compares the real-time distance-height curve with the pre-stored ideal reference curve to calculate the deformation value, and if the deformation value exceeds the deformation threshold, the door body movement is paused and a deformation warning is generated. If not, the state monitoring module is executed, and the variable value is transmitted to the state monitoring module. The state monitoring module extracts features based on real-time acquired vibration information and pressure information to obtain vibration characteristic values and pressure characteristic values, and transmits them to the intelligent decision module. The intelligent decision module quantifies the deformation value, vibration characteristic value and pressure characteristic value through a linear formula to obtain a running index. If the running index is greater than the running threshold, a running maintenance warning is generated. If not, no operation is needed, and the system returns to a low-power standby state, waiting for the next running instruction trigger.

[0029] In the description of the present specification, the description of the terms "one embodiment", "example", "specific example" and the like means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the invention. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0030] The preferred embodiments of the above disclosed invention are only used to help explain the invention. The preferred embodiments do not describe all the details and do not limit the invention to the specific embodiments described. Obviously, many modifications and changes can be made according to the content of the present specification. The present specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the invention, so that those skilled in the art can well understand and utilize the invention.

Claims

1. A method for intelligent monitoring of a garbage unloading roller shutter door based on multi-source sensor fusion, characterized in that, The method comprises the following steps: Step 1: When the roller shutter door receives an opening or closing instruction to start running, the door body image, vibration signal and pressure signal are collected; Step 2: Based on the foreign matter detection method combining background modeling and deep learning, it is judged whether there is an obstacle on the running path of the door body. If foreign matter is detected, the door body movement is paused and an alarm is issued. If not, step 3 is executed; Step 3: The door body deformation is monitored by an infrared distance sensor array. The real-time distance-height curve is compared with the pre-stored ideal reference curve to calculate the deformation variable value. If the deformation variable value exceeds the deformation threshold, the door body movement is paused and a deformation warning is generated; If not, step 4 is executed and the variable value is transmitted to step 5; Step 4: Based on the real-time acquired vibration information and pressure information, feature extraction is performed respectively to obtain vibration feature values and pressure feature values, which are transmitted to step 5; Step 5: The deformation variable value, vibration feature value and pressure feature value are quantified by a linear formula to obtain a running index. If the running index is greater than the running threshold, a running maintenance warning is generated. If not, no operation is needed and the system returns to a low-power standby state, waiting for the next running instruction. 2.The multi-source sensor fusion based intelligent monitoring method for garbage unloading roller shutter door according to claim 1, characterized in that, Foreign matter detection method combining background modeling and deep learning: A background model is pre-set. A stereo image is extracted. The absolute difference between the door body image and the background model is calculated in real time to obtain the difference value of each pixel. Then the difference values of each pixel are combined to form a difference image (x, y); The difference image (x, y) is binarized: if the difference value is greater than the pre-set threshold, the pixel coordinates of the difference image corresponding to the difference value are assigned a value of 1, otherwise a value of 0; then an open operation is performed on the binarized difference image (x, y) to identify the white connected region as a candidate foreign matter region; a trained target monitoring neural network is used to search and identify the door body image and the candidate foreign matter region and output the foreign matter region boundary box and confidence; A confidence threshold is pre-set. If the confidence is greater than the pre-set confidence threshold, a foreign matter warning is generated and the roller shutter door movement is paused. 3.The multi-source sensor fusion based intelligent monitoring method for garbage unloading roller shutter door according to claim 2, characterized in that, Background model construction process: The background model refers to a static reference picture representing the normal appearance of the door body region in the absence of foreign matter. In the absence of foreign matter, a number of images are collected to establish a background model. The color or grayscale value of each pixel in the image sequence is statistically analyzed to obtain a background model image with the same size as the real-time image, in which each pixel stores the color or grayscale value.

4. The multi-source sensor fusion-based intelligent monitoring method for garbage unloading roller shutter door according to claim 3, characterized in that, Comparison of real-time distance-height curve with pre-stored ideal reference curve to calculate deformation variable value: 4-1: On both sides of the roller shutter door opening, a number of measuring points are arranged at equal intervals from top to bottom. Each measuring point is equipped with an infrared distance sensor. Each sensor continuously measures the actual distance from the sensor to the door body surface at a fixed sampling frequency, denoted as di, where i is the index of the infrared distance sensor, i.e. the index of the measuring point. Thus, the actual curve of the actual distance di of each sensor i with the change of the door body height is denoted as di(H), where H is the door body height. 4-2, preset each sensor i respectively corresponds to a normal state distance Di changes with the door body height H ideal curve; 4-3, for each sensor i, extract its corresponding ideal curve, and the current actual curve and ideal curve are compared to get the deformation value; 4-4, thus each time the roller shutter door body from one end to the other end of the corresponding deformation value, the preset deformation threshold, if the deformation value is greater than the deformation threshold, the roller shutter door movement is suspended, and the roller shutter door deformation warning is generated.

5. The multi-source sensor fusion based intelligent monitoring method for garbage unloading roller shutter door according to claim 4, characterized in that, The coincidence comparison obtains the deformation value: Extract any one of the door body height Hj, calculate the absolute deviation Δdj of the two curves at the door body height Hj, Δdj=|di(Hj)-Di(Hj)|, j is any one door body height index, j=1, 2, 3, …, N, N is a positive integer, N represents the total number of different door body heights; thus the deviation sequence {Δdj} can be obtained; The root mean square error R is calculated by using the formula The Pearson correlation coefficient r between the ideal curve and the actual curve is calculated by taking the two variable sequences as two variable sequences, and the specific formula is as follows: , wherein is the distance mean of the actual curve at different door body heights, represents the distance mean of the ideal curve at different door body heights. The root mean square error R and the correlation coefficient r are calculated by the formula The deformation value is calculated by the formula Rmax is the preset maximum allowable root mean square error, and λ1 and λ2 are weight coefficients of the root mean square error and the correlation coefficient, respectively.

6. The multi-source sensor fusion-based intelligent monitoring method for garbage unloading roller shutter door according to claim 5, characterized in that, Vibration characteristic value extraction process: The vibration signal is extracted and denoted as A(t), and it is plotted as a vibration time-domain spectrum, where t is the time index and A(t) represents the amplitude at time t; the vibration time-domain spectrum of the vibration signal A(t) is analyzed: (1) by formula Calculate the root mean square value of the vibration signal , which characterizes the overall energy level of the vibration signal, and T represents the total time; (2) through the formula Calculate the waveform value of the vibration time-domain spectrum It represents the proportion of the impact component in the vibration signal relative to the overall average level; The root mean square value of the vibration time-domain spectrum and waveform value pass The nonlinear enhancement model calculates vibration characteristic values ​​to measure the vibration state of the roller shutter door during its movement, where η1 and η2 are the root mean square values, respectively. and waveform value The weighting coefficients.

7. The multi-source sensor fusion based intelligent monitoring method for garbage unloading roller shutter door according to claim 6, characterized in that, Pressure characteristic value extraction process: The extracted pressure signal is denoted as Yk, where k is the index of the pressure sensor at the bottom sealing strip of the door, k=1,2,3……K; for the pressure signal Yk at the same time t, it is obtained through the formula... The uniformity of pressure distribution is quantified to obtain the coefficient of variation P(t) between pressure sensors, where Let be the average pressure of K sensors at time t; from this, the coefficient of variation P(t) at each time can be obtained, and a time-domain spectrum of variation can be constructed. Spectral analysis of the time-domain spectrum of variation is then performed: (1) Through the formula Calculate the root mean square value of the vibration signal , which characterizes the degree of continuous uneven pressure; (2) through the formula Calculate the waveform values ​​of the time-domain spectrogram of the variation. It characterizes the impact characteristics and stability of the pressure variation sequence; The root mean square value of the variation time domain spectrum and the waveform value are calculated by a nonlinear enhancement model to measure the vibration state in the movement process of the roller shutter, wherein η3 and η4 are weight coefficients of the root mean square value and the waveform value respectively.

8. The intelligent monitoring system for garbage unloading rolling shutter door based on multi-source sensing fusion, characterized in that Applied to the intelligent monitoring system for garbage unloading roller shutter door based on multi-source sensor fusion according to any one of claims 1-7, the system comprises: a perception module, a foreign matter monitoring module, a deformation monitoring module, a state monitoring module and an intelligent decision module; The perception module is used for collecting door body images, vibration signals and pressure signals when the roller shutter door receives an opening or closing instruction to start running; The foreign matter monitoring module is based on a foreign matter detection method combining background modeling and deep learning, which judges whether there is an obstacle on the door body running path. If foreign matter is detected, the door body movement is suspended and an alarm is issued; if not, the deformation monitoring module is executed; The deformation monitoring module monitors the door body deformation through an infrared distance sensor array, compares the real-time distance-height curve with the pre-stored ideal reference curve to calculate the deformation value, and suspends the door body running and generates a deformation warning if the deformation value exceeds the deformation threshold; if not, the state monitoring module is executed, and the variable value is transmitted to the state monitoring module; The state monitoring module extracts features based on the real-time acquired vibration information and pressure information to obtain vibration characteristic values and pressure characteristic values, and transmits them to the intelligent decision module; The intelligent decision module comprehensively quantifies the running index by linear formula based on the deformation value, vibration characteristic value and pressure characteristic value, generates a running maintenance warning if the running index is greater than the running threshold, and does not need to perform any operation if not, and returns to the low-power standby state, waiting for the next running instruction trigger.

Images